Prismos AI

Prismos AI is a desktop assistant for people working with private documents and technical files. It uses local Ollama models to answer questions and retain knowledge for later conversations.

• Documents and memory: Chat with PDF, DOCX, PPTX and XLSX files, then explore saved concepts and their connections in a local knowledge graph; optional conflict alerts flag contradictory passages.
• Security analysis: Turn logs into incident timelines and findings mapped to MITRE ATT&CK, or review SSH, nginx, Docker and Kubernetes configurations for problems and suggested fixes.
• File and scene generation: Create reports, slide decks, small HTML apps and animated 3D scenes that open in a browser and work offline.
• Spreadsheets and recordings: Analyze CSV and XLSX data with calculations performed outside the model, generate standalone charts, and transcribe recordings locally with whisper.cpp and a downloaded speech model.
• Answer provenance: Optional receipts record which model and source material contributed to a reply, with verification available on the issuing machine.

It's free, open source under the MIT license, and runs on macOS, Windows and Linux without an account or telemetry. Chat works offline once models are installed. Private inference goes to Ollama on the same machine; optional web research, IMAP email and Yahoo Finance integrations use the network and are off by default. The app doesn't enforce offline behavior for Ollama or an external browser.

Local knowledge lives in an unencrypted SQLite database. Installers are unsigned, and the project hasn't undergone an independent security audit.